THE IMPACT OF MOBILE NUMBER PORTABILITY ON THE DIFFUSION OF MOBILE

TELECOMMUNICATIONS ACROSS EUROPE.

MACIEJ SOBOLEWSKI

A BSTRACT. This paper investigates the determinants of diffusion of mobile telecommunications in Euro-
pean Union. In addition to several technological and competitive related factors that are typically consid-
ered in other diffusion studies of mobile telephony present paper focuses on the impact of mobile number
portability (MNP) as a potential driver of diffusion process. Within the frames of logistic diffusion model
the study confirms the significant positive impact of MNP introduction on the speed of diffusion, however
the strength of this relation differs both with respect to type of subscriber contract and between EU-15 and
EU-12 countries.

1. I NTRODUCTION

There is a consensus in the literature that the development of mobile telecommunications has been
influenced by several events that have occurred throughout in the industry.
Main body of research on this issue was done when mobile penetration rate in most European markets
did not exceed 40-50%. In this early phase of industry development technological developments were
found to be the most important determinants of diffusion as shown by Gruber and Verboven [5], Gruber
[3] and [4]. Their main conclusion is that the transformation from analog to digital transmission and
utilization of higher frequency spectrum was the main driver of diffusion in the 90ties as it removed
strong capacity constraints. These papers also conclude that regulatory policy aimed at entry promotion
influenced the development of mobile telephony to a lesser extent. These conclusions are not surprising
since in that time action undertaken by national regulatory authorities to promote effective competition
between operators were limited mainly to licensing policy and did not utilized other measures.
Like in the case of many other innovations, the development of cellular telephony is a nonlinear
process. In almost all European countries diffusion path of mobile services resembles the sigmoid curve.
At present this process entered in a maturity phase. The penetration of digital mobile telephony exceeds
100%. In many countries there are already 3G networks under full coverage operation, offering next
generation services based on high speed data transfer. Competition between network operators in each
country continuously increases, putting a strong pressure on price levels. Beside ongoing regulation
within common regulatory framework, in last few years we have observed several competition enchaning
initiatives undertaken at national level and by European Commission. The most notable are promotion of
market entry by virtual operators (MVNO), mandated introduction of mobile number portability (MNP)
and recent fixing of price ceiling on roaming inside EU.
There is a strong supposition that some of these relatively recent events might influence the devel-
opment of mobile telecommunications in a very much similar way as technological transformation did
in the beginning of previous decade. These factors, which obviously could not have been considered in

Key words and phrases. diffusion models, mobile number portability.

Faculty of Economic Sciences, Warsaw University.
1
2 MACIEJ SOBOLEWSKI

earlier research, are potential drivers in the mature stage of diffusion process. The present study focuses
on one of those potential drivers, namely number portability. Our main research question is whether the
diffusion process of mobile telephony in the recent years has been speeded up by the introduction of
MNP.
The main motivation to post this research question is the ongoing debate about the effectiveness of
mobile number portability. In official documents by EC, OECD and national regulatory authorities as
well as consultation agencies, introduction of MNP was seen ex ante as the key action supporting the
development of competition. See for instance consultation document by Ovum [14] for Oftel and assess-
ment report by OECD [13].1 This view has been supported by the results of academic research.2
Despite the common belief that the MNP is an effective tool to promote competition, post implemen-
tation opinions are somehow ambiguous. While some research from US market and on EU level show
moderately positive impact of mobile number portability, it seems that on country level this regulatory
measure did not contributed in many cases to the growth of competitiveness, due to unfavorable condi-
tions for customers porting their numbers between network operators.3 Author believes that examination
of the relationship between MNP and the diffusion process of mobile telephony would be a valuable
contribution to this discussion.
The rest of the paper is organized in the following way. In the second section we briefly discuss
literature related to diffusion concept and mobile number portability. In the third section we turn to
model description. The fourth section is devoted to estimation details. Section fifth concludes. All
estimation results and figures are gathered up in the Annex.

2. R ELATED LITERATURE

2.1. Diffusion. According to the classical definition by Rogers [15] diffusion of innovation is a process
by which new ideas, products and technologies spread in the social system. The key questions in diffusion
analysis are in what manner and in what pace customers adopt new services and ideas. The process of
diffusion has four main elements: innovation, transmission channels, time and adoption dynamics and
social system.
Out of these four elements the one that is common for almost all innovations is the dynamics of diffu-
sion. Cumulated number of adoptions set against passing time graphically resembles sigmoid curve. In
the beginning phase, only few members of social system adopt the innovation, but the speed of adoptions
is constantly increasing over time. In a certain point of time - called the inflection point - the pace of
adoption slows down and the diffusion path flattens approaching the horizontal asymptote. In this matu-
rity phase the social system becomes satiated with the innovation. This kind of diffusion was confirmed
for thousands of products and services in hundreds of empirical research papers.
There are many modelling alternatives in diffusion analysis. The review of most commonly used is
provided by Mahajan and Peterson in [11]. As for mobile telecommunications the most popular models
are logistic curve, Gompertz curve, Bass model and non-uniform influence (NUI) model which is a
1
In 2002 OECD assessed that one of the most important shortcomings of regulatory environment of polish telecommuni-
cations market is lack of mobile number portability and its introduction was the basic recommendation for promoting price
competition, which was insufficient in light of OECD basket benchmarks.
2
For theoretic research the reader is refereed to papers by Klemperer [7], [8], [9]. For positive empirical results on the
relationship between number portability and price competition see Viard [16], Lee et al. [10] and Grzybowski [6].
3
This includes not only porting fee, but also a long transition period when sim module is disabled from operation. Porting
conditions differ to great extent among EU member states.
 NUMBER PORTABILITY AND DIFFUSION 3

generalization of the former. These models, which may also include covariates, have different number
of parameters and different properties concerning symmetry of diffusion curve and the range in which
inflection point may occur. The choice between alternative approaches in specific case/market depends
usually on the characteristics of empirical diffusion path such as its shape and maturity.

2.2. Switching costs. Without number portability a subscriber who wants to change his network opera-
tor has to give up his phone number. This is a classical example of switching cost in telecommunications.
According to Padila et al. [12] switching costs can be defined as real or perceived costs that are incurred
when changing supplier but which are not incurred by remaining with the current supplier. Switching
costs is a widely spread phenomenon which fundamentally changes the way in which firms behave and
markets operate. The existence of such costs leads to economies of scale in repeat purchasing, because a
customer who has previously bought from one firm incurs extra cost in purchasing an otherwise identical
product from a new firm, even if that product is sold at the same unit price. As a consequence, in markets
with consumer switching costs demand is less elastic and consumers have limited incentives to migrate
to cheaper offers which dampens competition.
In telecommunications lack of number portability is not the only instance of consumer switching
costs. Other forms of switching costs include simlocking handsets, fines for breaking contract and loyalty
discounts.
The impact of switching costs on market competition has been analyzed by many researchers. Surely
one of the most important contributions have been made by Klemperer in [8] and [7] and [9]. He discuses
the impact of switching costs on prices, market shares, profits and entry deterrence. The fundamental way
in which switching costs change firms behavior is that each supplier with an installed base of customers
will have interest in exploiting his customers, because they are locked-in by previous purchase. This
motivation, especially in case of no price discrimination, will cause a weak interest in competing for
rivals or new customers.
Another consequence of switching costs and lock-in mechanism is that for each supplier market share
is a fundamental goal of business strategy. Firms will invest in future market shares, since installed
customer base is valuable for them. In the dynamic market perspective, when firms first compete for new
customers who later on face switching costs of changing supplier, prices are lower in the beginning and
higher in later periods of competition as compared to identical market with no switching costs. This is
known as bargain then rip-off pricing. Although in this pricing pattern low prices for new customers are
followed by high prices for locked-in customers, but average market prices and profits are still higher
than in the absence of switching costs.

2.3. Mobile number portability. Although MNP was not mandated by the EU regulatory framework
before the year 2003, it was already offered in ten member states. Among UE countries, MNP was first
introduced in United Kingdom in January 1999. The next were: Spain (10/2000), Holland (01/2001),
Denmark and Portugal (07/2001), Sweden (09/2001), Italy (04/2002), Belgium (10/2002), Ireland and
Germany (11/2002). The change in legal framework making mobile number portability compulsory was
set in the European Parliament and Commission Directive 2002/22/EC on Universal Service and Users
Rights. The new framework took effect on 25 July 2003 and this date was the final deadline for mobile
network operators in all member states to have completed a full commercial launch of mobile number
portability.4
4This deadline was actually met in all EU-15 states, see Table 4.
4 MACIEJ SOBOLEWSKI

The effect of MNP introduction should be consistent with lowering switching costs. The main expec-
tation is that average market prices should fall down. A price decrease will affect not only old mobile
subscribers but also new customers and this should positively affect adoption rate. This is however only
true if new consumers have rational expectations. Naive customers do not see the relation between lower
prices today and higher prices tomorrow, so that reduction of switching costs would actually discourage
them from buying today because of less attractive promotional offers (weaker bargain part of pricing
strategy).
There are several types of benefits from MNP introduction. In a report [14] made by Ovum for
Oftel prior to MNP launch in United Kingdom, the benefits were estimated on average at 1.5 for
individuals and 99 for business subscribers. In case of business subscribers keeping their number
allows to avoid cost of informing clients (27); cost of updating advertisements, business cards, cars
repainting (174); cost of losing and substituting part of business contacts. On average 10 percent of
clients are lost and cost of substituting one client was estimated at 39 for small and medium enterprisers.
Although the results of cost benefit analysis for UK market were estimated positive, MNP has not to be
always welfare enhancing, as Buehler et al. argue in [2]. They identify different types of beneficiaries
of MNP introduction and point out on the negative consequences of this functionality on fixed-to-mobile
termination charges resulting from network identification problem.
The last part of literature includes few empirical studies of the impact of MNP on price competition.
In this respect there are at least two notable papers. The first one concerns the market for 800 toll-free
info lines in US (Viard [16]) and the second concerns mobile market in European Union (Grzybowski
[6]). There are also many empirical studies concerning the impact of lowering switching costs on market
price indices in other industries, such as airline, banking, and gasoline.5

3. M ODEL

The present paper utilizes logistic model of diffusion in a much similar way as Gruber in [4]. Therefore
we follow his notation in large parts. The logistic model can be specified as:

Yt 1
(1) = where b R+
M 1 + exp(a bt)

where a is location parameter and b is related to growth rate of diffusion process. Parameter M is
market potential. Greater values of location parameter shift diffusion curve backwards, so that the adop-
tion process can be described as more advanced as is depicted in Figure 1. For a complete description of
diffusion path within a framework of diffusion model all three parameters need to be estimated. How-
ever the market potential parameter M is difficult to estimate unbiased together with the remaining two,
especially in cases when diffusion curve has not reached its inflection point.6 In the present paper market
potential has been estimated simultaneously with location and speed parameters in a logistic model and it
turned to be underestimated for many countries. For this reason market potential has been also estimated
with three different models and the most reliable one out four estimates was selected.7

5Refer to Borenstein [1] as a illustrative example.

6See Gruber [4] for further discussion and solutions.
7They are put in column with header combined in tables 2a and 2b. For further details refer to section 3.1.
 NUMBER PORTABILITY AND DIFFUSION 5

For estimation purposes the logistic model given in (1) is often presented in the following linearized
form:
Yt
(2) zt = LN [ ] = a + bt
M Yt
Both location and speed parameters are assumed to be the following functions of explanatory variables
(covariates):

X
J X
K
(3) a = 0 + j D j + k xk
j=1 k=1

X
J X
K
b = 0 + j Dj + k xk
j=1 k=1

where Dj is dummy variable for introducing certain events in the regulatory environment and x is a
vector of other non-binary covariates affecting one or both parameters of diffusion.

3.1. Market potential. The three additional diffusion models used for market potential estimation were:
Gompertz model:

Yt
(4) = exp[exp(p(t q))]
M
Bass model:

Yt 1 exp[t(p + q)]
(5) =
M 1 + pq exp[t(p + q)]

and Non-Uniform Influence (NUI) model:

dYt Yt
(6) yt = = [p + q( )c ][M Yt ]
dt M
First two models have three parameters just like logistic model. The last model is a generalization
of Bass model. It has one more parameter and does not have a closed-form solution. The rationale
for utilization of these models is that they poses different mathematical properties allowing for flexible
estimation of market potential. The results of modelling market potential M for postpaid services and
for all types of contract are presented in Tables 2a and 2b.8
Figure 3 and first two columns of Table 4 present the results of logistic model estimation for all EU-
27 member states. Figure 3 reveals strong negative relationship between location and speed effects in
the sample. The least advanced countries with respect to mobile services diffusion (low value of a)
have at the same time greater values of speed parameter. This is well known symptom of international
convergence in mobile telecommunications between leaders and catching up countries.

8All estimations were done using nonlinear least squares procedure.

6 MACIEJ SOBOLEWSKI

4. E STIMATION

4.1. Description of Data and Hypotheses. The annual data covers the period from 1985 to 2006 for
all 27 EU member states. The data comes mainly from ITU World Telecommunication ICT Indicators
2007 database. We have also utilized two other sources of public data: www.gsmworld.com and Imple-
mentation Reports by the European Commission. List of variables taken into consideration included:9
sim postpaid - number of postpaid subscribers.
sim prepaid - number of prepaid subscribers.
gdp usd pc - gross domestic product per capita in USD. This variable is expected to have positive
impact on diffusion.
fixed per 100inh - number of main lines per 100 inhabitants. This variable might have positive or neg-
ative impact on diffusion depending whether fixed lines are used in a complement (ADSL) or substitute
(voice calls) manner to mobile services.
mnp intro - year of MNP introduction. This variable is expected to have positive impact on diffusion
as was justified in the literature section.
3G intro - year of 3G technology introduction.
2G operators - number of 2G operators.
2G intro - year of 2G technology introduction.
prepaid intro - year of prepaid introduction.
The last four variables are also expected to positively influence diffusion. While introduction of 2G
technology removed tight capacity constraints, the 3G technology offers new services which extend the
functionality of mobile subscription and therefore should attract new customers. The number of 2G
operators is a proxy for competition intensity. The prepaid mobile offer attracted very many customers,
who have either low income or specific usage profile or preferred not to bind with the network operator
for a long period of time. Due to its huge popularity prepaid offer boosted mobile penetration rate in
almost every country. As a illustration see Figure 5 for UK case. In some countries in reaction to prepaid
introduction a temporal decline in penetration of postpaid offer was observed.

4.2. Econometric specification and estimation results. The estimated model is based on equations (2)
and (3) and has the following form:

Yit
(7) zit = LN [ ] = ait + bit t + it where
Mi Yit
ait = f {intercept ; f ixed per 100inhit ; mnp introit ; 2G introit ; 3G introit ; 2G operatorsit ;
gdp usd pcit ; prepaid introit } and
bit = f {f ixed per 100inhit ; mnp introit ; 2G introit ; 3G introit ; 2G operatorsit ;
gdp usd pcit ; prepaid introit }

From the above equation it follows that all dependent variables entered formulas for both diffusion
parameters. This equation was estimated using panel regression procedure with fixed effects for six

9Selected country characteristics are presented in Table 4.

 NUMBER PORTABILITY AND DIFFUSION 7

different data sets. The panel has been limited to period 1993-2006 in order to obtain balanced data for
all 27 cross sections.10
First two estimations were done for all EU-27 countries on two different dependent variables: postpaid
subscribers and all subscribers regardless of type of contract.11 The reason for this separation was the
expectation that the MNP maybe attractive much more for the postpaid subscribers then for prepaid
group. The former group includes business subscribers, who are usually targeted by network operators
with different tariff schemes offered on the contract basis and are likely to have the greatest incentive to
change service provider. Therefore MNP might occur to be important for diffusion of mobile services
only in the postpaid subscribers population.
Another four estimations were performed on the same two dependent variables but separately for EU-
15 and EU-12 new member states. The reason for this separation is similar as before. In majority of new
member states the introduction of MNP was done only very recently, so that it might be difficult to detect
its impact on diffusion of mobile subscriptions.
The results of all panel regressions are presented in Tables 6 and 7. Table 6 indicates that in the EU-
27 the diffusion of all types of mobile subscriptions has been positively influenced by introduction of
mobile number portability, prepaid offer, 2G technology, and also to the lesser extent by the number of
2G operators. Negative estimates for speed effects indicate that the positive shocks on location parameter
are diminishing over time. Mobile number portability seems to be very important driver of diffusion in
EU-12 countries while in EU-15 introduction of 2G technology and prepaid offer played a major role.
Competition effect although significant play a minor role in both groups of countries.
While MNP has been insignificant for diffusion of all mobile subscriptions in EU-15 countries it is
important for diffusion of postpaid contracts in this group, however again introduction of 2G technology
is the most influential factor (see Table 7). For both dependent variables, 3G introduction has no impact
on the diffusion in EU-12.

5. C ONCLUSIONS

The study yields interesting results concerning the impact of mobile number portability on the dif-
fusion of mobile telecommunications. Number portability influenced the diffusion of all mobile sub-
scriptions in EU-27 and in new member states (EU-12). It has also been important factor of postpaid
subscriptions diffusion in EU-15. The latter result is the most significant, because in EU-15 number
portability and prepaid services were introduced almost in the same time and analyzing only the postpaid
segment enabled to separate the impact of both.
For both types of mobile subscriptions 2G remains the most important diffusion driver in EU-15
countries, while in EU-12 this factor has not been important. This not surprising since new member
states have launched mobile services when analog technology was passing away. It seems that in case of
EU-12 number portability and prepaid plays similar role as main drivers of diffusion as 2G technology
introduction in EU-15 in the 90ties.
Introduction of 3G technology occurred to be insignificant for diffusion in any configuration and
competition intensity measured by number of 2G operators occurred to have positive, although very little
10The alternative estimation method used in Gruber [4] and Gruber and Verboven [5] - nonlinear least squares on pooled

data did not provide interesting and significant results. Also contrary to both papers, no restrictions on parameters values for
variables entering both location and speed equations occurred to be significant.
11The variable prepaid intro has been ruled out from all estimations on postpaid data.
8 MACIEJ SOBOLEWSKI

impact on diffusion of mobile telecommunications. With respect to the last result and also to the huge
role of 2G introduction in EU-15 the present study is consistent with results of previous research.

R EFERENCES
1. Borenstein Severin, Selling Costs and Switching Costs: Explaining Retail Gasoline Margins, RAND Journal of Economics
22 (1991), no. 3, 354369.
2. Buehler Stefan, Haucap Justus, Mobile Number Portability, Journal of Industry, Competition and Trade 4 (2004), no. 3.
3. Gruber Harald, Competition and Innovation. The Diffusion of Mobile Telecommunications in Central and Eastern Europe,
Economics and Policy 13 (2001).
4. , The Economics of Mobile Telecommunications, 1 ed., Cambridge University Press, 2005.
5. Gruber Harald, Frank Verboven, The Diffusion of Mobile Telecommunications Services in the European Union, European
Economic Review 45 (2001).
6. Grzybowski Lukasz, Regulation of Mobile Telephony accross the European Union: An Empirical Analysis, Journal of
Regulatory Economics 28 (2005), no. 1.
7. Klemperer Paul, Markets with Consumer Switching Costs, Quarterly Journal of Economics 102 (1987), 375394.
8. , The Competitiveness of Markets with Switching Costs, RAND Journal of Economics 18 (1987), no. 1, 139150.
9. , Competition when Consumers have Switching Costs: An Overview with Aplications to Industrial Organization,
Macroeconomics, and International Trade, Review of Economic Studies 62 (1995), 515539.
10. Lee Sang-Woo, Kim J. Dan, Park Myeong-Cheol, Demand for Number Portability in the Korean Mobile Telecommuni-
cations Market: Contingent Valuation Approach, Proceedings of the 37th Hawaii International Conference on Systems
Sciences (2004).
11. Mahajan Vijajy, Robert Peterson, Models for Innovation Diffusion, SAGE Publications, 1985.
12. NERA, Switching Costs, Annexe A - Literature Review, OFT Economic Discusion Paper 5, Office of Fair Trading, April
2003, Autorzy: Padilla Atiliano Jorge, Wiliams Mark, McSorley Ciara, Fernandez Daniel, Reyes Tatiana.
13. OECD, Regulatory Reform in the Telecommunications Industry in Poland. From Transition to New Regulatory Challenges,
OECD Reviews of Regulatory Reform, March 2002.
14. OVUM, Economic Evaluation of Number Portability in the UK Mobile Telephony Market, Report for Oftel, Office of
Telecommunications, July 1997.
15. Rogers Everett, Diffusion of Innovations, Free Press, 1983.
16. Viard V. Brian, Do Switching Costs Make Markets More or Less Competitive?: The Case of 800-Number Portability,
Research Paper Series. Stanford University - Graduate School of Business (2004), no. Paper No. 1773(R1).

WARSAW U NIVERSITY, FACULTY OF E CONOMIC S CIENCES , UL . D LUGA 44/50, 00-241 WARSAW, P OLAND
E-mail address: maciej.sobolewski@uw.edu.pl
The impact of mobile number portability on the diffusion of mobile
telecommunications across Europe.

Annex.

Figure 1. Stylized diffusion paths in logistic model.

1,2 1,2
a=-1; b=0.5
1 a=-2; b=0.5 1
a=-2; b=1
a=-2; b=0.5 with shock on "a"
0,8 0,8
diffusion

0,6 0,6

0,4 0,4

0,2 0,2

0 0

time

Table 2a. Market Potential. Prepaid and postpaid mobile.

For Spain the Gompertz curve estimation did not converge. In case of underestimation by all
four models, market potential was calculated as 1,05 of current sim-to-population ratio.
Currently in EU-27 sim-to-population ratio equals on average 1,059.

A1
Table 2b. Market Potential. Postpaid mobile.

(-) indicates no convergence. In case of underestimation by all four models, market potential
was calculated as 1,05 of current sim-to-population. Currently in EU-27 sim-to-population
ratio in postpaid equals on average 0,44.

Figure 3. Location and speed effects. Logistic diffusion model on postpaid

and prepaid mobile.

1,4

1,2

1
speed effects "b"

0,8

0,6

0,4

0,2

0
-14 -12 -10 -8 -6 -4 -2 0 2
-0,2
location effects "a"

A2
Table 4. Location a and speed effects b from logistic model of
diffusion and selected characteristics of mobile industries.

Figure 5. MNP and prepaid introduction. UK case.

A3
 Table 6. Diffusion of postpaid and prepaid mobile. Panel estimation.

EU-27 EU-15* EU-12 (new member states)

Variable Estimate St. Error Pr > |t| Estimate St. Error Pr > |t| Estimate St. Error Pr > |t|

Location effects

FIXED_PER_100INH 0.073024 0.0131 <.0001 0.121662 0.0250 <.0001 0.074476 0.0175 <.0001

MNP_INTRO 1.455606 0.6689 0.0303 0.838975 0.7366 0.2563 11.19149 2.9567 0.0002

_2G_INTRO 0.651572 0.3366 0.0538 4.813074 2.5812 0.0640 0.952448 0.6293 0.1326

_3G_INTRO 0.165831 0.6374 0.7949 1.118492 0.8727 0.2018 -0.48144 1.9239 0.8028

_2G_OPERATORS 0.415969 0.1338 0.0020 0.39844 0.1866 0.0342 0.603855 0.2513 0.0177

GDP_USD_PC -0.00007 0.000031 0.0178 -0.00008 0.000038 0.0260 -0.00036 0.000099 0.0004

PREPAID_INTRO 0.84536 0.3118 0.0071 1.636682 0.4976 0.0012 2.093542 0.6595 0.0019

Speed effects

FIXED_PER_100INH -0.00388 0.000770 <.0001 -0.00505 0.00138 0.0003 -0.01024 0.00150 <.0001

MNP_INTRO -0.09043 0.0417 0.0310 -0.04488 0.0443 0.3126 -0.74516 0.1945 0.0002

_2G_INTRO -0.10104 0.0435 0.0209 -0.5349 0.2892 0.0662 -0.20112 0.0977 0.0416

_3G_INTRO -0.02118 0.0387 0.5843 -0.06303 0.0484 0.1946 0.030367 0.1348 0.8221

_2G_OPERATORS -0.01021 0.00950 0.2834 -0.00933 0.0118 0.4306 -0.03655 0.0251 0.1484

GDP_USD_PC 3.556E-6 1.288E-6 0.0061 4.045E-6 1.623E-6 0.0136 0.000027 5.534E-6 <.0001

PREPAID_INTRO -0.06307 0.0256 0.0142 -0.11116 0.0349 0.0017 -0.25359 0.0755 0.0010

(*) h0 in F test saying DF= MSE= R-Sq.= DF= MSE= R-Sq.= DF= MSE= R-Sq.=
no fixed effects is 320 0,237 0,970 166 0,198 0,967 127 0,190 0,982
always rejected.

A4
 Table 7. Diffusion of postpaid mobile. Panel estimation for period 1993-
2006.

EU-27 EU-15 EU-12 (new member states)

Variable Estimate St. Error Pr > |t| Estimate St. Error Pr > |t| Estimate St. Error Pr > |t|

Location effects

FIXED_PER_100INH 0.054379 0.0159 0.0007 -0.01331 0.0245 0.5870 0.045564 0.0243 0.0636

MNP_INTRO 0.195732 0.8171 0.8108 1.390151 0.7274 0.0577 3.257542 4.1183 0.4304

_2G_INTRO 0.912258 0.4099 0.0268 7.497437 2.5274 0.0035 -0.67294 0.8772 0.4444

_3G_INTRO 1.293135 0.7781 0.0975 0.617212 0.8548 0.4713 0.700301 2.6776 0.7941

_2G_OPERATORS 0.23244 0.1629 0.1545 0.401162 0.1837 0.0304 0.6342 0.3501 0.0725

GDP_USD_PC -0.00008 0.000038 0.0347 -0.00008 0.000037 0.0324 -0.00022 0.000138 0.1114

Speed effects

FIXED_PER_100INH -0.00466 0.000934 <.0001 0.001257 0.00135 0.3533 -0.01165 0.00209 <.0001

MNP_INTRO -0.0114 0.0509 0.8231 -0.08901 0.0438 0.0435 -0.176 0.2710 0.5172

_2G_INTRO -0.19238 0.0528 0.0003 -0.89529 0.2832 0.0019 0.034212 0.1363 0.8022

_3G_INTRO -0.07059 0.0472 0.1356 -0.05658 0.0474 0.2344 0.022149 0.1875 0.9062

_2G_OPERATORS -0.00564 0.0115 0.6250 -0.01931 0.0116 0.0975 -0.0608 0.0353 0.0877

GDP_USD_PC 3.713E-6 1.571E-6 0.0187 3.283E-6 1.597E-6 0.0413 9.87E-6 7.792E-6 0.2077

(*) h0 in F test saying DF= MSE= R-Sq.= DF= MSE= R-Sq.= DF= MSE= R-Sq.=
no fixed effects is 315 0,349 0,940 163 0,186 0,951 125 0,367 0,958
always rejected.

A5